Quantitative customer survey results and other forms of customer feedback may provide insight into a customer's level of satisfaction with a particular company's services. Especially in instances where a customer contacts a service department over the phone and receives customer support. However, receiving feedback service scores from a customer as quantized levels of satisfaction (1—less satisfied, 2, 3, 4 and 5—highly satisfied, etc.) leaves a large level of uncertainty as to what each customer really likes or dislikes about a particular company, product and/or service.
Today, analyzing the data associated with a customer call is mostly a manual procedure. This can be burdensome and difficult to analyze as the recorded components of a customer call and how the customer ranked various services via a survey score are not easy to review and understand in a reasonable period of time. Data analysts are required to review the survey data and identify the individual customer and their respective account. Next, the data analyst must also access data in remote locations to listen to the recorded call (if available) to identify the trouble areas of the call. Once the call and/or comment-based recordings are made available, most analysis is done manually, on an ad-hoc level. Even advanced audio mining does not provide a concise and real-time analysis of the customer's true customer service experience.
If data analysis is performed without a corresponding audio mining application then samplings of calls must be listened to individually. This leaves different data analysts with the responsibility of making statements and decisions about the entire population of customers based on various call recordings. Such a task is usually the only situation for survey comment analysis. Any data analysis must be performed using a separate tool and linked back manually to the macro-level customer survey data, if linked back at all, in an effort to be efficient. In such instances where speech analytics are used, the entire process is performed in disparate systems, which is long and burdensome and far from a real-time analysis.